Six governments and ten AI companies signed the Paris Call for AI safety and security on Monday, November 9, 2026 — and unlike most summit statements, it names specific things developers are committing to do: publish evaluation findings in a standard format, accept independent third-party evaluation, and build fail-safe mechanisms that can shut down model operations, including during training. Why it matters: the signatories include Anthropic, OpenAI and Microsoft, so the pledges cover a large share of the frontier models people actually use.
November 10, 2026
What happened
The joint declaration, published by the Prime Minister of Canada’s office, was signed by the heads of state and government of Bahrain, Canada, Finland, France, Norway and the Netherlands. The company signatories are Anthropic, Black Forest Labs, Cohere, Domyn, Hugging Face, Microsoft, Mistral AI, OpenAI, Sakana AI and Synthesia. Four civil society groups also signed: LawZero, SaferAI, AI Safety Connect and Ifri. The statement remains open to new endorsements.
The declaration frames frontier models as inherently global — developed, deployed and felt across borders — and says no single state, company or international organization can manage their risks alone. It builds on earlier efforts it names directly, including the Hiroshima process, the Frontier Safety Commitments of the Seoul AI Summit, the G7 Evian AI meeting and a United Nations Security Council meeting on AI held on September 23.
The three commitments
First, the signatories commit to a common framework for evaluating and managing the risks of frontier models, supported by shared technical standards, tools and methods. The stated focus is dangerous misuse, including in the cyber and CBRN domains, and risks from growing autonomy or misalignment, including loss of control. They also agree to publish the non-sensitive findings of their evaluations in a standardized format and to create a publicly accessible shared repository that could include red-teaming protocols, benchmarking tools, risk assessments and stress-test results.
Second, they affirm responsibility and accountability across the AI ecosystem. Developers commit to disclose model capabilities, limitations and known risks, to support evaluation by qualified independent evaluators where appropriate, and to implement fail-safe mechanisms to shut down model operations within their control, including during training and internal use. The text also acknowledges the risk from removing or circumventing safeguards — including abliteration — while preserving the benefits of open-weight development.
Third, they commit to information sharing, reporting and early warning: rapid alert systems, aligned incident classification, and timely sharing of information on serious cyber and AI control incidents between companies and governments. Policymakers also commit to secure, confidential channels for whistleblowers to raise concerns about systemic risks without fear of retaliation.
Why it matters
Declarations are only as good as their follow-through, and this one is voluntary. But the specifics matter for anyone buying or building on frontier models: standardized, published evaluation results would make it far easier to compare models on safety as well as capability, and a shared incident-reporting system would mean a serious failure at one lab becomes a warning for everyone else, not a rumor. The shutdown pledge is also notable timing — it lands the same day OpenAI described its own internal kill-switch system. Our take: treat this as a floor being negotiated in public, not a finished rulebook — and watch whether the promised shared repository and standardized evaluation format actually appear.
Frequently asked questions
Who signed the Paris Call for AI safety?
Six governments — Bahrain, Canada, Finland, France, Norway and the Netherlands — plus ten companies including Anthropic, OpenAI, Microsoft, Mistral AI, Cohere and Hugging Face, and four civil society organizations.
Is the Paris Call legally binding?
No. It is a joint declaration of commitments, not a treaty or law. Its force comes from public accountability and from the national frameworks the signatories say they will build on it.
What is the fail-safe shutdown commitment?
Frontier developers commit to implement mechanisms that can shut down model operations within their control — including restrictions, pauses or shutdowns during training and internal use.
Sources: Prime Minister of Canada (joint declaration, primary text), Canada Newswire